NeoFox - NEOantigen Feature tOolboX

NeoFox annotates neoantigen candidate sequences with published neoantigen features.

For a detailed documentation, please check out https://neofox.readthedocs.io

If you use NeoFox, please cite the following publication:
Franziska Lang, Pablo Riesgo-Ferreiro, Martin Löwer, Ugur Sahin, Barbara Schrörs, NeoFox: annotating neoantigen candidates with neoantigen features, Bioinformatics, Volume 37, Issue 22, 15 November 2021, Pages 4246–4247, https://doi.org/10.1093/bioinformatics/btab344

Table of Contents

1 Implemented neoantigen features
2 NeoFox requirements
3 Usage from the command line
4 Input data
5 Output data

1 Implemented Neoantigen Features

NeoFox covers the following neoantigen features and prediction algorithms:

§ currently not supported for mouse

2 NeoFox Requirements

NeoFox depends on the following tools:

• Python 3.11
• BLAST 2.10.1
• netMHCpan 4.1
• netMHCIIpan 4.3
• MixMHCpred 2.2 (optional)
• MixMHC2pred 2.0.2 (optional)
• PRIME 2.0 (optional)

Install from PyPI:

pip install neofox

Or install from bioconda:

conda install bioconda::neofox

3 Usage from the command line

NeoFox can be used from the command line as shown below or programmatically (see https://neofox.readthedocs.io for more information).

neofox --input-file neoantigens_candidates.tsv \
    --patient-data patient_data.txt \
    --output-folder /path/to/out \
    [--output-prefix out_prefix]  \
    [--organism human|mouse]  \
    [--rank-mhci-threshold 2.0] \
    [--rank-mhcii-threshold 5.0] \
    [--num-cpus] \
    [--config] \
    [--patient-id] \
    [--with-all-neoepitopes] \
    [--verbose]

• --input-file: tab-separated values table with neoantigen candidates represented by long mutated peptide sequences as described here (extensions .txt and .tsv) or JSON file neoantigens in NeoFox model format as described here (extension .json)
• --patient-data: a table of tab separated values containing metadata on the patient as described here
• --output-folder: path to the folder to which the output files should be written
• --output-prefix: prefix for the output files (optional)
• --with-all-neoepitopes: output annotations for all MHC-I and MHC-II neoepitopes on all HLA alleles (optional)
• --rank-mhci-threshold: MHC-I epitopes with a netMHCpan predicted rank greater than or equal than this threshold will be filtered out (optional)
• --rank-mhcii-threshold: MHC-II epitopes with a netMHCIIpan predicted rank greater than or equal than this threshold will be filtered out (optional)
• --organism: the organism to which the data corresponds. Possible values: [human, mouse]. Default value: human
• --num-cpus: number of CPUs to use (optional)
• --config: a config file with the paths to dependencies as shown below (optional)
• --patient-id: patient identifier (optional, this is only relevant if the column patientIdentifier is missing in the candidate input file)
• --verbose: get detailed logs

The optional config file with the paths to the dependencies can look like this:

NEOFOX_REFERENCE_FOLDER=path/to/reference/folder
NEOFOX_BLASTP=path/to/ncbi-blast-2.10.1+/bin/blastp
NEOFOX_NETMHCPAN=path/to/netMHCpan-4.1/netMHCpan
NEOFOX_NETMHC2PAN=path/to/netMHCIIpan-4.3/netMHCIIpan
NEOFOX_MIXMHCPRED=path/to/MixMHCpred-2.2/MixMHCpred
NEOFOX_MIXMHC2PRED=path/to/MixMHC2pred-2.0.1/MixMHC2pred_unix
NEOFOX_MAKEBLASTDB=path/to/ncbi-blast-2.8.1+/bin/makeblastdb
NEOFOX_PRIME=/path/to/PRIME-2.0/PRIME

4 Input data

4.1 Neoantigen candidates in tabular format

This is an dummy example of a table with neoantigen candidates:

where:

• gene: the HGNC gene symbol
• mutatedXmer: the neoantigen candidate sequence, i.e. the mutated amino acid sequence. The mutation should be located in the middle, flanked by 13 amino acid on both sites (IUPAC 1 respecting casing, eg: A)
• wildTypeXmer: the equivalent non-mutated amino acid sequence (IUPAC 1 respecting casing, eg: A)
• patientIdentifier: the patient identifier
• rnaExpression: RNA expression. (optional) (see NOTE) This value can be in any format chosen by the user (e.g. TPM, RPKM) but it is recommended to be consistent for data that should be compared.
• rnaVariantAlleleFrequency: the variant allele frequency calculated from the RNA (optional, this will be estimated using the dnaVariantAlleleFrequency if not available)
• dnaVariantAlleleFrequency: the variant allele frequency calculated from the DNA (optional)

NOTE: If rnaExpression is not provided, expression will be estimated by gene expression in a respective TCGA cohort and this value will be used for relevant features. The TCGA cohort to be used for imputation of gene expression needs to be indicated in the tumorType in the patient data (see below). If tumorType is not provided, expression will not be imputed.

4.2 Neoantigen candidates in JSON format

Besides tabular format, neoantigen candidates can be provided as a list of neoantigen models in JSON format as shown below. To simplify, only one full neoantigen model is shown:

[{
    "patientIdentifier": "Ptx",
    "gene": "BRCA2",
    "mutation": {
        "wildTypeXmer": "AAAAAAAAAAAAALAAAAAAAAAAAAA",
        "mutatedXmer": "AAAAAAAAAAAAAFAAAAAAAAAAAAA"
    }
}]

4.3 Patient-data format

This is an dummy example of a patient file:

where:

• identifier: the patient identifier
• mhcIAlleles: comma separated MHC I alleles of the patient for HLA-A, HLA-B and HLA-C. If homozygous, the allele should be added twice.
• mhcIIAlleles: comma separated MHC II alleles of the patient for HLA-DRB1, HLA-DQA1, HLA-DQB1, HLA-DPA1 and HLA-DPB1. If homozygous, the allele should be added twice.
• tumorType: tumour entity in TCGA study abbreviation format (https://gdc.cancer.gov/resources-tcga-users/tcga-code-tables/tcga-study-abbreviations). This field is required for expression imputation. The supported tumor types are listed under "Input data" in the documentation.

5 Output data

The output data is returned by default in tsv and json format. With the command line flag --with-all-neoepitopes, two additional files are generated containing the epitope candidates for MHCI and MHCII with NetMHCpan predictions below the given thresholds.